| 1. |
|
|
| 2. |
- Bengtsson Bernander, Karl
(författare)
-
Equivariant Neural Networks for Biomedical Image Analysis
- 2024
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- While artificial intelligence and deep learning have revolutionized many fields in the last decade, one of the key drivers has been access to data. This is especially true in biomedical image analysis where expert annotated data is hard to come by. The combination of Convolutional Neural Networks (CNNs) with data augmentation has proven successful in increasing the amount of training data at the cost of overfitting. In this thesis, equivariant neural networks have been used to extend the equivariant properties of CNNs to more transformations than translations. The networks have been trained and evaluated on biomedical image datasets, including bright-field microscopy images of cytological samples indicating oral cancer, and transmission electron microscopy images of virus samples. By designing the networks to be equivariant to e.g. rotations, it is shown that the need for data augmentation is reduced, that less overfitting occurs, and that convergence during training is faster. Furthermore, equivariant neural networks are more data efficient than CNNs, as demonstrated by scaling laws. These benefits are not present in all problem settings and which benefits will occur is somewhat unpredictable. We have identified that the results to some extent depend on architectures, hyperparameters and datasets. Further research may broaden the performed studies to explain how the results occur with new theory.
|
|
| 3. |
- Bengtsson Bernander, Karl, et al.
(författare)
-
Replacing data augmentation with rotation-equivariant CNNs in image-based classification of oral cancer
- 2021
-
Ingår i: Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications. - Cham : Springer International Publishing. - 9783030934194 - 9783030934200 ; , s. 24-33
-
Konferensbidrag (refereegranskat)abstract
- We present how replacing convolutional neural networks with a rotation-equivariant counterpart can be used to reduce the amount of training images needed for classification of whether a cell is cancerous or not. Our hypothesis is that data augmentation schemes by rotation can be replaced, thereby increasing weight sharing and reducing overfitting. The dataset at hand consists of single cell images. We have balanced a subset of almost 9.000 images from healthy patients and patients diagnosed with cancer. Results show that classification accuracy is improved and overfitting reduced if compared to an ordinary convolutional neural network. The results are encouraging and thereby an advancing step towards making screening of patients widely used for the application of oral cancer.
|
|
| 4. |
- Blache, Ludovic, et al.
(författare)
-
SoftCut: : A Virtual Planning Tool for Soft Tissue Resection on CT Images
- 2018
-
Ingår i: Medical Image Understanding and Analysis. - Cham : Springer. - 9783319959207 ; , s. 299-310
-
Konferensbidrag (refereegranskat)abstract
- With the increasing use of three-dimensional (3D) models and Computer Aided Design (CAD) in the medical domain, virtual surgical planning is now frequently used. Most of the current solutions focus on bone surgical operations. However, for head and neck oncologic resection, soft tissue ablation and reconstruction are common operations. In this paper, we propose a method to provide a fast and efficient estimation of shape and dimensions of soft tissue resections. Our approach takes advantage of a simple sketch-based interface which allows the user to paint the contour of the resection on a patient specific 3D model reconstructed from a computed tomography (CT) scan. The volume is then virtually cut and carved following this pattern. From the outline of the resection defined on the skin surface as a closed curve, we can identify which areas of the skin are inside or outside this shape. We then use distance transforms to identify the soft tissue voxels which are closer from the inside of this shape. Thus, we can propagate the shape of the resection inside the soft tissue layers of the volume. We demonstrate the usefulness of the method on patient specific CT data.
|
|
| 5. |
|
|
| 6. |
- Christersson, Albert, et al.
(författare)
-
Comparison of 2D radiography and a semi-automatic CT-based 3D method for measuring change in dorsal angulation over time in distal radius fractures
- 2016
-
Ingår i: Skeletal Radiology. - : Springer Science and Business Media LLC. - 0364-2348 .- 1432-2161. ; 45:6, s. 763-769
-
Tidskriftsartikel (refereegranskat)abstract
- Objective The aim of the present study was to compare the reliability and agreement between a computer tomography-based method (CT) and digitalised 2D radiographs (XR) when measuring change in dorsal angulation over time in distal radius fractures. Materials and methods Radiographs from 33 distal radius fractures treated with external fixation were retrospectively analysed. All fractures had been examined using both XR and CT at six times over 6 months postoperatively. The changes in dorsal angulation between the first reference images and the following examinations in every patient were calculated from 133 follow-up measurements by two assessors and repeated at two different time points. The measurements were analysed using Bland-Altman plots, comparing intra- and inter-observer agreement within and between XR and CT. Results The mean differences in intra- and inter-observer measurements for XR, CT, and between XR and CT were close to zero, implying equal validity. The average intra- and inter-observer limits of agreement for XR, CT, and between XR and CT were +/- 4.4 degrees, +/- 1.9 degrees and +/- 6.8 degrees respectively. Conclusions For scientific purpose, the reliability of XR seems unacceptably low when measuring changes in dorsal angulation in distal radius fractures, whereas the reliability for the semi-automatic CT-based method was higher and is therefore preferable when a more precise method is requested.
|
|
| 7. |
- Khonsari, Roman H., et al.
(författare)
-
Fronto-facial advancement and bipartition in Crouzon-Pfeiffer and Apert syndromes : Impact of fronto-facial surgery upon orbital and airway parameters in FGFR2 syndromes
- 2016
-
Ingår i: Journal of Cranio-Maxillofacial Surgery. - : Elsevier BV. - 1010-5182 .- 1878-4119. ; 44:10, s. 1567-1575
-
Tidskriftsartikel (refereegranskat)abstract
- A major concern in FGFR2 craniofaciosynostosis is oculo-orbital disproportion, such that orbital malformation provides poor accommodation and support for the orbital contents and peri-orbita, leading to insufficient eyelid closure, corneal exposure and eventually to functional visual impairment. Fronto-facial monobloc osteotomy followed by distraction osteogenesis aims to correct midfacial growth deficiencies in Crouzon–Pfeiffer syndrome patients. Fronto-facial bipartition osteotomy followed by distraction is a procedure of choice in Apert syndrome patients. These procedures modify the shape and volume of the orbit and tend to correct oculo-orbital disproportion. Little is known about the detailed 3D shape of the orbital phenotype in CPS and AS, and about how this is modified by fronto-facial surgery.Twenty-eight patients with CMS, 13 patients with AS and 40 control patients were included. CT scans were performed before and after fronto-facial surgery. Late post-operative scans were available for the Crouzon–Pfeiffer syndrome group. Orbital morphology was investigated using conventional three-dimensional cephalometry and shape analysis after mesh-based segmentation of the orbital contents.We characterized the 3D morphology of CPS and AS orbits and showed how orbital shape is modified by surgery. We showed that monobloc-distraction in CPS and bipartition-distraction in AS specifically address the morphological characteristics of the two syndromes.
|
|
| 8. |
- Khonsari, R H, et al.
(författare)
-
Shape and volume of craniofacial cavities in intentional skull deformations
- 2013
-
Ingår i: American Journal of Physical Anthropology. - : Wiley. - 0002-9483 .- 1096-8644. ; 151:1, s. 110-119
-
Tidskriftsartikel (refereegranskat)abstract
- Intentional cranial deformations (ICD) have been observed worldwide but are especially prevalent in preColombian cultures. The purpose of this study was to assess the consequences of ICD on three cranial cavities (intracranial cavity, orbits, and maxillary sinuses) and on cranial vault thickness, in order to screen for morphological changes due to the external constraints exerted by the deformation device. We acquired CT-scans for 39 deformed and 19 control skulls. We studied the thickness of the skull vault using qualitative and quantitative methods. We computed the volumes of the orbits, of the maxillary sinuses, and of the intracranial cavity using haptic-aided semi-automatic segmentation. We finally defined 3D distances and angles within orbits and maxillary sinuses based on 27 anatomical landmarks and measured these features on the 58 skulls. Our results show specific bone thickness patterns in some types of ICD, with localized thinning in regions subjected to increased pressure and thickening in other regions. Our findings confirm that volumes of the cranial cavities are not affected by ICDs but that the shapes of the orbits and of the maxillary sinuses are modified in circumferential deformations. We conclude that ICDs can modify the shape of the cranial cavities and the thickness of their walls but conserve their volumes. These results provide new insights into the morphological effects associated with ICDs and call for similar investigations in subjects with deformational plagiocephalies and craniosynostoses.
|
|
| 9. |
|
|
| 10. |
- Malmberg, Filip, et al.
(författare)
-
Relaxed Image Foresting Transforms for Interactive Volume Image Segmentation
- 2010
-
Ingår i: MEDICAL IMAGING 2010. - : SPIE. - 9780819480248 ; 7623
-
Konferensbidrag (refereegranskat)abstract
- The image Foresting (IFT) is a framework for image partitioning, commonly used for interactive segmentation. Given an image where a subset of the image elements (seed-points) have been assigned correct segmentation labels, the IFT completes the labeling by computing minimal cost paths from all image elements to the seed-points. Each image element is then given the same label as the closest seed-point. Here, we propose the relaxed IFT (RIFT). This modified version of the IFT features an additional parameter to control the smoothness of the segmentation boundary. The RIFT yields more intuitive segmentation results in the presence of noise and weak edges, while maintaining a low computational complexity. We show an application of the method to the refinement of manual segmentations of a thoracolumbar muscle in magnetic resonance images. The performed study shows that the refined segmentations are qualitatively similar to the manual segmentations, while intra-user variations are reduced by more than 50%.
|
|
